Skip to content

README.md

Latest commit

 

History

History
99 lines (76 loc) · 3.31 KB

README.md

File metadata and controls

99 lines (76 loc) · 3.31 KB
Raw

VTL Engine

Maven Central

Architecture Pattern

The engine follows a clear 4-step flow for dataset operators:

  1. visitors/*FunctionsVisitor (and ClauseVisitor) only dispatch parse-tree nodes.
  2. semantics/<domain>/*Executor applies VTL semantics (roles, structure, validation, planning).
  3. ProcessingEngine executes mechanical operations (join, filter, aggregate, etc.).
  4. semantics.DatasetResults.withStructure(...) re-attaches VTL structural metadata to mechanical results.

This keeps concerns separated: visitors route, semantics decide, processors execute.

The engine never depends on a concrete runtime: semantics call the ProcessingEngine interface from vtl-model. Implementations are plugged in at bootstrap via ProcessingEngineFactory.

Diagram

flowchart TB
  subgraph antlr["vtl-antlr"]
    PT["ANTLR parse tree"]
  end

  subgraph engine["vtl-engine"]
    subgraph visitors["visitors — dispatch"]
      FV["*FunctionsVisitor / ClauseVisitor"]
      EV["expression/*Visitor"]
    end

    subgraph semantics["semantics — VTL meaning"]
      EX["&lt;domain&gt;/*Executor"]
      ATTR["attribute/ — viral propagation"]
      DR["DatasetResults.withStructure"]
    end

    subgraph scalar["functions/providers — scalar natives"]
      TF["NumericFunctionsProvider, …"]
    end

    subgraph processors["processors — default runtime"]
      IMPE["InMemoryProcessingEngine"]
    end
  end

  subgraph model["vtl-model — contracts &amp; data"]
    PE(("ProcessingEngine"))
    PEF["ProcessingEngineFactory"]
    DE["DatasetExpression"]
    RE["ResolvableExpression"]
    DS["Dataset · Structured · Role"]
  end

  subgraph spark["vtl-spark · vtl-spark4 — distributed runtime"]
    SPE["SparkProcessingEngine"]
  end

  PT --> FV
  PT --> EV

  FV --> EX
  EX --> ATTR
  EX --> PE
  PE --> IMPE
  PE --> SPE
  IMPE --> DE
  SPE --> DE
  EX --> DR
  DR --> DE

  EV --> RE
  EV --> TF
  RE --> DS
  DE --> DS

  PEF -.->|"bootstrap"| IMPE
  PEF -.->|"bootstrap"| SPE

  classDef iface fill:#e8f4fc,stroke:#4a90d9
  class PE iface
Loading

Dataset path (operators): visitor → executor → ProcessingEngine → DatasetResults.withStructure → DatasetExpression.

Scalar path (row expressions): expression/*Visitor builds ResolvableExpression directly — no ProcessingEngine.

Package Roles

Module Package / class Role
vtl-antlr generated visitors VTL grammar, parse tree
vtl-model ProcessingEngine, DatasetExpression, ResolvableExpression, Dataset Runtime-agnostic contracts and data bindings
vtl-engine visitors Parse-tree dispatch only
vtl-engine semantics.<domain> VTL operator logic (join, aggregation, validation, time, …)
vtl-engine semantics.attribute Cross-cutting viral attribute propagation
vtl-engine processors.InMemoryProcessingEngine Default in-memory ProcessingEngine
vtl-engine functions.providers Built-in scalar native functions (Map<String, List<Method>>)
vtl-spark / vtl-spark4 SparkProcessingEngine Spark-backed ProcessingEngine